Fooling Computer Vision
Data Skeptic
English - January 22, 2020 18:38 - 25 minutes - 23.3 MB - ★★★★★ - 477 ratingsScience Technology machinelearning datamining datascience science skepticism statistics Homepage Download Apple Podcasts Google Podcasts Overcast Castro Pocket Casts RSS feed
Previous Episode: Algorithmic Fairness
Next Episode: Interpretable One Shot Learning
Wiebe van Ranst joins us to talk about a project in which specially designed printed images can fool a computer vision system, preventing it from identifying a person. Their attack targets the popular YOLO2 pre-trained image recognition model, and thus, is likely to be widely applicable.